Aerospace Instrument-Making Annotation << Back APPLICATION OF NEURAL NETWORKS FOR CALCULATING THE FUNCTIONING OF LEVER-FLOAT VALVES IN AVIATION SYSTEMS D.O. Pushkarev, D.Y. Kiselev The article analyzes the possibilities of using neural networks in the design, production and testing of lever-float valves. The paper shows the advantages of neural networks over classical calculation methods. The structure of neural networks and a neuron is disassembled, the principle of operation of a neural network is shown. The disadvantages of existing calculation methods are described. The principle of calculating the lever-float valves has been analyzed (a lever-float valve with four elements was used as an example), the main details that need to be considered in the calculation are shown. The principle of constructing a neural network is considered, it is shown how the analysis of statistical data and the selection of parameters for training a neural network is performed. The work shows how the parameters of the product interact with each other. A comparative analysis was performed between the test data and the results calculated using a neural network. Recommendations are given on how this neural network can be applied in the future, as well as directions for improving the existing neural network model. Keywords: neural network; levers; theory of mechanisms and machines; test; machine learning; neuron. DOI: 10.25791/aviakosmos.5.2021.1218 Pp. 19-27.